Cal Poly
Analog Hardware Implementation of a Linearly Constrained Quadratic Program Real-Time Solver
Abstract
dc:description.abstractThis thesis presents the design, implementation, and analysis of a hardware system for solving Linearly Constrained Quadratic Programs (LCQPs) in real time. The architecture follows a generalized feedback structure composed of three key elements: gradient descent on the quadratic cost function, saturation-based nonlinearity to enforce inequality constraints, and an integral controller with an anti-windup mechanism to regulate dynamic behavior and determine steady-state error. This majority analog system converges with equilibria that satisfy the Karush-Kuhn-Tucker (KKT) optimality conditions. Using a representative LCQP, this work presents simulation of the circuit in PLECS and LT Spice to confirm the feasibility of the novel architecture presented. In addition, a hardware implementation was designed on a printed circuit board (PCB) and tested in terms of accuracy and convergence speed, achieving equilibrium values within 0.5% error of the theoretical solutions and at speeds that rivals other hardware prototypes of its kind. These promising results lay the groundwork for the next phase of development: a VLSI-based implementation aimed at achieving even higher levels of integration and computational efficiency. Ultimately, this work contributes to the growing body of research on analog computing architectures that synthesize control, optimization, and computation in physical hardware.
Degree
thesis:*- Name thesis:degree_name
- MS in Electrical Engineering
- Discipline thesis:degree_discipline
- Electrical Engineering
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Tylutki, Claire E
- Tylutki, Claire
- Contributors dc:contributor
-
- Jason Poon
- Electrical Engineering
- College of Engineering
Subjects
dc:subject × 5Identifiers
dc:identifier.*- Repository record dc:identifier
- https://digitalcommons.calpoly.edu/theses/3010